Crystal structure of bovine angiogenin at 1.5-A resolution

Proc. Natl. Acad. Sci. USA

Vol. 92, pp. 2949-2953, March 1995 Biochemistry

Crystal structure of bovine angiogenin at 1.5-A resolution K. RAvI ACHARYA*t, ROBERT SHAPIROt§, JAMES F. RIORDANt, AND BERT L. VALLEEI *School of Biology and Biochemistry, University of Bath, Claverton Down, Bath BA2 7AY, United Kingdom; and *Center for Biochemical and Biophysical Sciences and Medicine and §Department of Pathology, Harvard Medical School, 250 Longwood Avenue, Boston, MA 02115

Contributed by Bert L. Vallee, December 27, 1994

ABSTRACT The capacity of angiogenin (Ang) to induce blood vessel growth is critically dependent on its ribonucleolytic activity. Crystallography and mutagenesis of human Ang have previously shown that its pyrimidine binding site is obstructed by Gln-117, implying that a conformational change is a key part of the mechanism of Ang action. The 1.5-A-resolution crystal structure of bovine Ang, in which glutamic acid is substituted for Gln-117, now confirms that a blocked active site is characteristic of these proteins. Indeed, the inactive conformation of bovine Ang is stabilized by a more extensive set of interactions than is that of human Ang. The three-dimensional structure of the putative receptor binding site is also well conserved in the two proteins. The Arg-Gly-Asp segment of this site in bovine Ang, which is replaced by Arg-Glu-Asn in human Ang, does not have a conformation typical of an integrin recognition site.

same angiogenic potency, and, in general, similar enzymatic properties (see below). However, it has two differences in primary structure of potential significance: Glu-118 replaces Gln-117, and it contains a putative Arg-Gly-Asp (RGD) recognition element not present in hAng. We here report that Glu-118 in bAng obstructs the pyrimidine binding site in precisely the same manner as does Gln-117 in hAng, and is, in fact, involved in a more extensive set of stabilizing interactions than its counterpart in hAng. Furthermore, the RGD sequence is not structurally similar to known integrin-dependent recognition sites. Importantly, the bAng structure has been determined to 1.5-A resolution, which defines other interactions pertinent to the architecture of the active site in greater detail. It also provides a clearer view of additional functionally important regions, including the putative receptor binding site.

Angiogenin (Ang) is unique among angiogenic molecules in that it is a member of the pancreatic ribonuclease (RNase) superfamily (1) and, in fact, is a ribonucleolytic enzyme (2). Its enzymatic activity is extremely weak compared to that of the digestive RNases (2, 3) but is critical for its capacity to induce neovascularization (4, 5). Its in vivo substrate remains to be identified. Ang was first isolated from human tumor cell conditioned medium (6) and subsequently from normal human plasma (7). Angs have also been purified from other mammalian sera (8, 9) and from cow's milk (10). Most previous efforts to elucidate the structural basis for and relationship between enzymatic and angiogenic activities have focused on human Ang (hAng). Mutagenesis, proteolysis, and chemical modification studies have thus far identified several residues that play important or essential roles in catalysis or substrate binding and have at least for the enzymatic differences between Ang part#Haccounted and pancreatic RNase (4, 5, 11-13). Moreover, they have demonstrated that angiogenic activity requires not only an intact catalytic site but also another region of Ang, thought to constitute a cell binding site. Recently, hAng has been crystallized (14) and a 2.4-Aresolution structure has been determined (15), which provides a basis for the unique functional properties of the protein. Most strikingly, it reveals that the site corresponding to the pyrimidine binding pocket of RNase A is blocked by Gln-117. Mutations of this residue to Ala and Gly increase enzymatic activity (16), implying that this conformation for Ang also exists in solution. The structure of native Ang in which access to the binding site is blocked presumably must undergo extensive reorientation in order to allow binding and cleavage of RNA. This conformational change might also serve to activate Ang at its target site in vivo. In continuing efforts to further understand the structurefunction relationships of Ang, we have now determined the x-ray structure of bovine angiogenin (bAng). This protein (Mr 14,595) has 64% sequence identity to hAng (10), virtually the

METHODS bAng was isolated from unpasteurized milk essentially as described for its purification from plasma (8). Needle-shaped crystals (maximum dimensions, 0.7 mm x 0.3 mm x 0.3 mm) were grown at 16°C using the hanging drop method with a reservoir solution of 0.1 M sodium acetate/0.2 M ammonium acetate, pH 4.5/30% polyethylene glycol 4000. They belong to the orthorhombic space group P212121 (a = 30.69 A, b = 54.45 A, c = 74.75 A) with one molecule per asymmetric unit and 43% (vol/vol) solvent content. Diffraction data (AD-lab data set) were collected from native crystals (1.6 A) using a Siemens area detector mounted on a Siemens rotating anode x-ray source using CuKa radiation (50 kV, 80 mA). Overall, 1900 frames of data were collected (0.25° per frame; crystal to detector distance, 10 cm; 20 angle, 350; 120-130 sec per frame exposure). The data were processed using the XDS package (17). Higher-resolution data to 1.5 A (SRS data set) were collected on station 9.5 of the Synchrotron Radiation Source (SRS; Daresbury, U.K.), using an 18-cm (diameter) MARresearch imaging plate system with an x-ray beam of 0.88-A wavelength (0.2-mm collimator) from one crystal with 1.50 oscillations. The SRS data were processed using the MOSFLM package (A. Leslie, Medical Research Council, Cambridge, U.K.). The two data sets were scaled by using the program 3DSCALE (18) (Table 1). The structure was determined using the MERLOT molecular replacement programs (19) and the search was based on the hAng protein model (15) with bAng amino acid sequence changes incorporated into the model. The angles from rotation and translation function search gave a single, self-consistent set of translation vectors on the Harker sections. Cycles of positional and simulated annealing refinement using the X-PLOR package (20) and model building using FRODO (21) gave the refined model (using all data, 17,940 Abbreviations: Ang, angiogenin; bAng, bovine Ang; hAng, human Ang; SRS, Synchrotron Radiation Source. tTo whom reprint requests should be addressed. ¶The atomic coordinates have been deposited in the Protein Data Bank, Chemistry Department, Brookhaven National Laboratory, Upton, NY 11973 (entry code 1AGI). This information is embargoed for 1 year (coordinates) from the date of publication.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

2949

2950

Proc. NatL Acad ScL USA 92

Biochemistry: Acharya et al

(1995)

Table 1. bAng x-ray data collection statistics

Rsym,t % Data set Nm* % complete Nit 9.0 1.5-A native (SRS) 76,423 16,542 80 (A = 0.88 A) 94 4.6 1.6-A native (AD-lab) 73,455 16,060 (A = 1.542 A) 7.1 87 1.5-A native (all) 149,878 18,103 *Number of measurements. tNumber of independent reflections. *Rsym = Y2iyXhI(h) - Ii(h)I/Xij4i(h), where Is(h) is the ith measurement of reflection h and I(h) is the mean of the intensity. reflections, 8.0-1.5 A), which has an R factor (EIIF0I [FcjI/M[Foj) of 0.203 (0.201, 0.199, and 0.197 for F . o,, F> 2or, and F > 3cr, respectively; rms deviation of bond lengths from ideality, 0.008 A; bond angles, 1.53°). The structure contains 91 bound water molecules. All non-glycine residues lie within the allowed regions of the Ramachandran (4, qi) plot. Individual temperature factors were refined for each atom during the final stages of refinement (the mean temperature factor for protein atoms is 19.8 A2).

RESULTS AND DISCUSSION Overall Structure. In general, the electron density map is of high quality (Fig. 1) and the entire structure is well-defined except for the three N-terminal residues, the four C-terminal residues, the side chains of Asp-60 and Arg-61, and the terminal portions of the side chains of Lys-20, Lys-83, Arg-90, and Arg-102, all of which have high temperature factors and are disordered. The structure of bAng has the "RNase A fold" (Fig. 2) and is very similar to that of hAng (15). Fig. 3 shows the sequence alignment of bAng and hAng and indicates which bAng residues are either solvent-inaccessible or involved in crystal contacts. Compared to the human protein, bAng has single-residue extensions at both termini and no insertions or deletions. The rms difference between the C" positions of 121 equivalent residues in the two crystal structures (bAng 2-122) is 1.24 A. The regions that deviate

FIG. 2. Polypeptide fold of bAng, drawn with the program MOLSCRIPT (22). The secondary-structure elements shown [based on program DSSP (23)] contain the following residues: 4-15 (Hi), 23-34 (H2), 42-48 (B1), 50-59 (H3), 63-66 (B2), 70-74 (B3), 77-85 (B4), 94-102 (B5), 104-109 (B6), and 112-117 (B7).

most significantly are the N-terminal residues 2 and 3 (5.16 and 2.47 A, respectively) and the C-terminal residues 120-122 (1.65, 5.07, and 9.54 A, respectively). The difference in the C-terminal region is particularly noteworthy as discussed below. In hAng, residues 117-121 form a 310 helix that is absent in bAng; instead, the corresponding segment appears to be flexible and disordered after the second residue, Ser-119. The amino acid sequence in this segment is also poorly conserved between the two Angs. Exclusion of these residues and amino acids 2 and 3 from the structural comparison reduces the rms deviation in Cc positions for the two proteins to 0.51 A, indicating a high degree of similarity between the structures. Only five residues (21, 50, 68, 88, and 89) differ in position by >1.0 A. All but one of these (Asn-50) are on a loop. Two peptide bonds in the bAng crystal structure adopt a cis conformation: those connecting Arg-38 with Pro-39 and # ##

#

*

*

*

bAng a Q D d yR Y i H F L T Q H Y D A K P k G R n 23 hAng - Q D n sR Y t H F L T Q H Y D A K P q G R d 22 * + * * #+#**# ++**** * D e Y C f n m M k n R r L T r P C K D r N T F I H G 49 D r Y C e s i M r r R g L T s P C K D i N T F I H G 48

* # # #* * ** # ** +# # # N K n d I K A I C E d r N G q P y R g d L R I S K S 75 N K r s I K A I C E n k N G n P h R e n L R I S K S 74

* +* * **

+#

# ## ##

* #* *#

+

e F Q i T i-C K h k G G S s r P P C r Y g A T e d SF Q v T t C K 1 h G G S pw P P C q Y r A T a g *

** * *+# +

* * +* *+#

#

R vi V V g C E N G L P V H f D e S f i tp r h f R n V V V a C E N G L P V H 1 D q S i f r rp -

s

FIG. 1. Portion of a 21Fo1 - jFCI electron density map of bAng contoured at l.Ocr using the refined structure at 1.5-A resolution.

100 99 125 123

FIG. 3. Sequence alignment of bAng and hAng. Identities are indicated by uppercase letters. Hyphens indicate that there is no corresponding equivalent residue. The solvent-inaccessible residues [